MotionShine01 / Drop
Choose a photo
Drop it, pick it, or paste a screenshot.
Free tool
AI image extender, more room around a photo.
Turn a photo into a wide banner, a square post or a story without cropping it. Pick the new shape, drag the photo into place, and an open AI model paints the empty edges from what is already there. It works best on simple backgrounds such as sky, walls and studio backdrops. It runs in your browser, so nothing is uploaded.
New shape
Best on simple edges: sky, sea, walls, grass, a studio backdrop. People, faces and text cut by the edge are not finished well.
This runs in your browser. Your photos are never uploaded.
How it works
Three steps, nothing to install.
02 / Frame
Pick the shape and place it
Wide, square, story or larger. Drag the photo to where it should sit.
03 / Extend
Let the AI paint the edges
The empty parts are filled from the photo's own edges. Check them, then download.
Questions
Good to know.
Which AI model does it use?
LaMa (big-lama), an open model for filling holes in pictures, released under the Apache 2.0 licence, in an ONNX version by Carve. It is the same model as the object remover, so if you used that, it is already in your browser.
Is my photo uploaded?
No. The model is downloaded to your browser from Hugging Face, the open model library, the first time (208 MB), and the photo is worked on in this tab.
What does it do well, and what not?
LaMa continues what it sees at the edges: sky, clouds, water, walls, floors, grass, sand, plain or blurred backgrounds. It does not invent new things, so a person, a face, a building or text cut by the edge is not completed properly, and very large new areas get plainer the further they are from the photo. Leave a little space around the subject, and add only as much room as you need.
Is it like generative fill?
No. Generative fill in apps such as Photoshop uses much larger models that run on their servers and can draw new objects. This tool runs on your own computer with a small model that only continues the background, which is why it is free and private.
Does it keep the quality of my photo?
Your photo is laid back on top unchanged. The new parts are painted at up to 1,280 pixels across the new frame and enlarged to match, so on a big photo they can look softer than the photo up close.
How long does it take?
It works in steps from the edges outwards: a few for a small change, a dozen or more for a big one. With a graphics card (WebGPU) each step takes moments; without one it can take several seconds each.